Download Optimizing Data Acquisition for Automated de novo Sequencing

Optimizing Data Acquisition for Automated de novo Sequencing
1Bioinformatics
Iain Rogers1, Michaela Scigelova2, Gary Woffendin2
Purpose: We have evaluated various data acquisition methods on the LTQ Orbitrap to obtain data most suitable
for automated de novo sequencing. We then reviewed the examples where an integrated approach using de
novo sequencing and blast search indicated an amino acid substitution or sequence variation in analysed
peptides.
Methods: Medium complex peptide sample (6 proteins digest) was analysed by nano-LC-MS/MS using three
different methods combining various fragmentation modes (linear ion trap vs C-trap) and levels of accuracy of
fragment ion detection. De novo sequencing was performed with PEAKS Studio 4.2.
FIGURE 2. An overview of the acquisition methods used (inset). HCD spectra generated in the C-trap
delivered the highest number of complete correct de novo sequences from among the three methods
tested.
Precursor
Scan
Fragmentation
Method
HCD
Total 102
MS/MS
Detection
Orbitrap
CID (LTQ)
LTQ
Orbitrap
CID (LTQ)
Orbitrap
Orbitrap
HCD (C-Trap)
Orbitrap
20
1
Linear ion trap LTQ
C-trap
17
CID, OT
Detection
Total 49
CID, LT
Detection
Total 51
6
The results were filtered with ‘Xcorr vs charge state’ filter so as to ensure less than 3% false positive rate (FPR)
of identification established using reverse database searches.
These results were used to benchmark the outcome of de novo sequencing with PEAKS Studio 4.2. The mass
tolerance set for Orbitrap-measured spectra (both precursor and fragment ions) was 0.01 Da while 1 Da was
used for the ion trap-measured fragments. The result was considered correct if the complete sequence matched
the one obtained in the database search, while allowing for the following ambiguities: Leu = Ile; GlyGly = Asn. In
the particular case of fragment masses measured with the LTQ: Gln = Lys, and one possibility of an amino acid
positional interchange was still admissible, i.e. NA = AN.
FIGURE 3: Qualitative MS/MS spectral differences. CID spectra contain prominent y- and b-ion series, while
HCD spectra tend to show a dominant y-ion series and presence of immonium ions.
100
y4+1
CID MS/MS
515.4
90
Linear Ion Trap Detection
80
70
While the standard search with BioWorks was able to identify all six of the proteins in the sample, there remained
a considerable number of spectra (335 out of 638 total good quality MS/MS spectra) unassigned after a
database search. The identified proteins’ sequences and reversed sequences were used as the reference
database for SPIDER algorithm which is part of PEAKS software package.
87
53
50%
SPIDER homology search enabled us to assign further 85 MS/MS spectra which belonged to the 6 proteins
analysed. The variations in the sequence included modifications like deamidation (Figure 6).
FIGURE 6. Homology search with SPIDER highlighted many cases of peptide modification or amino acid
substitution, like this example of deamidation.
60%
272
981
489
40%
444
66
30%
De_Novo
Database
60
y6+1
50
y5+1
40
10
932.3
1031.4
233.0
+1 b11 +1
y 10
b7+1 b8+1
833.2
762.3
y7+1 y8+1
b5+1
303.2
b2+1
b3+1 b4 +1
346.1
300
516.3
b6+1
798.4
649.3
447.2
400
500
600
700
m/z
885.7
800
1101.5
1144.4
+1
y 11
y9+1
1214.6
1000.6
900
1000
1100
Relative Abundance
HCD MS/MS
Orbitrap Detection
70
60
a
1+
y41+
b21+
515.3292
2
233.0953
205.1007
25+
Results
The database search identified confidently (less than 3% FPR) 158, 108, and 167 spectra from the CID LTQ
Detection, CID Orbitrap Detection, and HCD Orbitrap Detection experiments, respectively. These spectra
were de novo sequenced with PEAKS.
The MS/MS spectra obtained with higher energy collisional dissociation (HCD) in the C-trap differed from those
generated by standard collision-induced dissociation (CID) in the ion trap (Figure 3).
Conclusions
De novo sequencing requires the best data quality available.
The LTQ Orbitrap method employing higher energy collisional dissociation (HCD) performed in the C-trap
generates data which is well suited for automated de novo sequencing.
The C-trap MS/MS spectra are detected in the Orbitrap analyser. The high resolution is very useful for confident
charge state assignment of fragment ions from higher charge state precursors (we observed correct sequence
assignment for several 4+ peptides).
HCD data are characterised by 3 ppm accurate mass measurement of fragment ions, prominent and mostly
complete y-ion fragment series, presence of low m/z fragments and immonium ions.
Success rate of automated de novo sequencing expressed as correct amino acid residue assignment was 91%
and increased to 98% for peptides up to 15 amino acid long.
Homology search using SPIDER in PEAKS helped increase protein sequence coverage and identified large
number of modifications/amino acid substitutions.
The data of this study were used for further improving the performance of PEAKS Studio 4.2 software with ion
trap MS/MS data. See poster xxx at this meeting.
y81+
885.5496
175.1187
y111+
y91+
y51+
1000.5753
614.3975
y31+
y11+
10
0
21 to 25
685.4350
80
20
16 to 20
Peptide Length
FIGURE 5. HCD spectra with fragments measured accurately in the orbitrap enable a confident assignment of
amino acids. This example shows peptide LLVTQTMK with internal glutamine residue that can not be mistaken
for a lysine even though the mass difference would be just 38 mDa.
1200
y61+
90
30
11 to 15
1272.5
1101.6243
100
40
<11
b12+1
y101+
50
0%
HCD data (167 MS/MS spectra) were further evaluated considering the assignment of individual amino acids to a
correct position within the peptide. In total, 2252 of 2475 amino acids were correctly assigned which represents
91% of all residues. The successful outcome of de novo sequencing was clearly dependent on the length of the
peptide. The program delivered a correct assignment for 98% of residues in peptides with 15 or less amino acids
(Figure 4).
b10 +1
685.4
614.4
y2+1
0
Orbitrap
b9 +1
416.3
DQTVLQNTDGDNNEAWAK
DQTVIQNTDGNNNEAWAK
10%
The results were first evaluated considering the number of complete correct sequences (Figure 2).
y3+1
20
Peptides in a six protein digest mix (Dionex; 100 fmol) were separated via Surveyor™ LC equipped with
MicroAS™ autosampler (Thermo Fisher Scientific) using a peptide trap column (PSDVB, Michrom
Bioresources) and a PicoFrit™ nanocolumn packed with 5 μm BioBasic™ C18, 300 Å pore size (New
Objective), flow rate 300 nL/min. A gradient of 5 - 40% acetonitrile in 40 minutes was used. The tandem mass
spectrometric analysis was performed on the LTQ Orbitrap (Thermo Fisher Scientific). The so called ‘Top3’
acquisition routine was used, comprising a full MS scan followed by 3 Data Dependent™ MS/MS scans. All full
scan spectra were recorded at resolving power 30,000 in the Orbitrap analyzer. The MS/MS spectra
(fragmentation and detection) were obtained as detailed in Figure 2 (inset).
The database search (BioWorks™ 3.3.1) was performed against the Uniprot-SwissProt database (containing
216,381 entries), considering fully tryptic peptides and carboxyamidation on Cys (+ 58.0055 Da) as a fixed
modification. The precursor mass tolerances for all experiments were 10 ppm; the fragment tolerances
considered for Orbitrap-measured MS/MS spectra were 10 mmu, whilst ion trap spectra were searched with 1
Da tolerances.
57
The high mass accuracy of the HCD fragmentation spectra eliminates de novo sequencing errors that result from
the fact that masses of certain combinations of amino acid residues are very similar in mass. For instance, it is
easy to resolve the difference between K and Q (128.172 x 128.130 Da; Figure 5), or MM and YV (262.081 and
262.131 Da).
20%
30
Methods
25
70%
0
Relative Abundance
ESI source
1
80%
37
28
Introduction
FIGURE 1. The diagram of LTQ Orbitrap mass spectrometer showing the key components: the linear ion trap
LTQ, C-trap, and Orbitrap mass analyser. These mass analysers can operate independently of each other or
in a convenient parallel mode, maximising the duty cycle. CID fragmentation is carried out in the linear ion
trap, and the fragments can be detected either in the linear ion trap or Orbitrap. HCD fragmentation is
performed in the C-trap and the fragments are detected in the Orbitrap (inset).
100%
90%
Results: Higher collisional energy dissociation in the C-trap provided the best quality data for automated de
novo sequencing (no low mass cutoff, fragments measured with low ppm accuracy). Several sequence
variations on known peptides were identified with the PEAKS sw, increasing protein coverage.
De novo sequencing enables identification of peptides and proteins from unsequenced genomes [1,2] or
validation of the results of a database search [3]. To be of practical use this process must be automated, with a
throughput matching that of data acquisition.
The LTQ Orbitrap™ delivers routine mass measurements with deviations of less than 3 ppm (external
calibration). In the context of proteomics experiments measuring the precursor ion highly accurately means
fewer false positive identifications [4, 5].
There is, however, no clear consensus regarding the benefit of MS/MS mass accuracy. This is because the
accurate mass detection in the Orbitrap analyser takes longer than the fragment detection in a linear ion trap,
resulting in potentially less peptides being fragmented and identified. Also, the LTQ Orbitrap can fragment
peptides in the linear ion trap or in the C-trap, each method being characterised by particular spectra qualities
(Figure 1).
We performed a detailed comparison of data acquisition methods on LTQ Orbitrap with respect to their suitability
for automated de novo sequencing with PEAKSTM Studio 4.2 software. As this package combines de novo
sequencing with BLAST searches in databases we were also interested in indications of amino acid substitutions
or unexpected modifications.
FIGURE 4. Effect of peptide length on the outcome of automated de novo sequencing. 98% residues were
correctly assigned for peptides 15 or fewer amino acids long.
% R es idue s
Overview
Solutions, Canada; 2Thermo Fisher Scientific, UK;
1214.7102
y71+
416.2605
y21+
303.1775
b31+
346.1793
200
300
400
500
600
700
m/z
800
References
Sunyaev et al., Anal. Chem., 2003, 75, 1307-15.
Frank et al., J. Proteome Res., 2005, 4, 1287-95.
Wielsch et al., J. Proteome Res., 2006, 5, 2448-56.
Yates, J.R. et al., Anal. Chem., 2006, 78, 493-500.
Zubarev, R., Mann, M., Mol. Cell. Proteomics, 2007, in press.
798.5185
900
1000
1100
1200
PicoFrit is a registered trademark of New Objective, Inc. SEQUEST are registered trademarks of the University of Washington. All other trademarks
are the property of Thermo Electron Corporation and its subsidiaries.